- cross-posted to:
- Aii@programming.dev
- cross-posted to:
- Aii@programming.dev
You must log in or register to comment.
running a model on a reasonably specced laptop
my 128GB M5 Max MacBook Pro
TIL a $7000 laptop is “reasonably specced”
You can run it on smaller hardware too, people managed to get 3.6 to run with 16 gb ram with stuff like quants and MTP. But the fact that you can run a model that performs as well as a frontier model from half a year ago on any laptop is frankly mind blowing.
im not bothered by it.
but also… i wouldn’t use it as my sole model to perform every task. [waves hand toward huggingface.] ( a gazillion bajillion models, specialized for all types of tasks).edit: i checked. and there aren’t as many models as i expected, for this task(didnt test them). and… small multimodal model failed badly(see picture of the iguana … or whatever the hell that is)
it’s a strange benchmark. but it seems reasonable to expect LLMs to be able to do this.
i tried using it as conversation agent for home assistant and it always calculated the dates wrong
A much more effective way to use models is to have them drive a deterministic tool like a calendar instead. Also, Qwen 3.8 is complete overkill for a task like that, you’re just wasting energy for no reason here. A much smaller model trained on agentic tool use would both work more reliably and use way less power.
I did make it use getTime every time but when I tell it to make a calendar entry for the “next Tuesday” for example it would be off by at least a week. Sometimes even making the entry in the past.
I’d agree with you on it being overkill, but I wrote SK many scripts and automations involving LLM tasks that rely on a capable model to determine the flow of a task that I heavily prefer the 27ban models
currently running gemma4 and for my use case it performs much better than qwen3.8
I find what the model was RL trained on is really important. It looks like Qwen 3.8 is mainly focused on agentic coding, so it does really well there. But once you throw it at tasks outside the training then things start to fall apart fast.
yeah the training data is super important, i just hoped that i would fare well inside the home assistant MCP setting :D
Taking your comment about it being overkill as food for thought, i just pulled the 12b version of gemma4 and run that now. I’ll observe if it works just as good for my usecase and that way i just saved like 10GB of VRAM :D
nice
